TWI343192B - Decoding method - Google Patents

Description

1343192

, :· TW5226PA VI. Description of the invention: t [Technical field to which the invention pertains] ^ The present invention relates to a decoding method. [Prior Art] • Refer to Figure 1 and Figure 2, where the 1st and 2nd 81st is the Huffman table of the coding end, and Figure 2 shows the reconstruction of the traditional decoder. In order to ensure data consistency, the encoder and decoder must use the same Huffman table. Therefore, the encoding end generates the symbol array Symm and the Huffman code number array NC[] of the same bit length according to the Huffman table. The symbol array Sym□ includes all the symbols of the Huffman table 1〇, and the code values corresponding to the Huffman codes corresponding thereto (c〇de ^ small and large, sequentially arrange the symbols corresponding to the code values, For example, the code value of the Huffman code of Huffman table is from 0 to 3, 3, 4, 5, 6, 12, 24, 48 96, 192, 384 (ie, the number of binary digits) , 〇1〇, 〇11,1〇〇, ι〇ι η/ 1110, 11110'm11〇'1111110,11111110,lllllul〇), so #the symbol in the symbol array is also in the code value order of the Huffman code Arranged, so the corresponding symbol array is Sym[]={〇, 1,2, 3, 4, 5, 6, 7, 8, 9, 1〇, 11}. The same bit το length of Hoffman The code number array NC□ records the number of Huffman codes corresponding to each bit length, and the Huffman code number array corresponding to the same bit length of the Huffman table 1^(:[卜丨〇 ,丨,5, 1,^ ,1' 1' 1' 0' 0' 0' 〇, 〇, 〇, 0}, which means that there are a total of Huffman codes of length i; There are five Huffman codes of length 3; Huffman codes of lengths of 4 to 9 each have i ; Huffman code length of 10 to 16 are encoded as a square array of symbols do called terminal] 343 192 3 inches.

The TW5226PA and the Huffman code array NC[] of the same bit length are stored in the File Header, and then the decoder end is based on the symbol array Symy of the head and Hoffman of the same bit length. The code number array Nc□ reconstructs the Huffman table 2〇, and the decoding end generates a lookup table based on the Huffman table 20.凊> Idea: The Huffman code index value of Figure 2 is exactly the same as its symbol value, and those symbols are just arranged in a continuous value. These are just special cases, not all Huffman tables have this situation.

Please refer to Fig. 2 and Fig. 3 at the same time. Figure 3 shows the lookup table of the Leading 1 s Search. Referring to Table 3, the lookup index value m' is equal to the number of leading i of the Huffman code in the Huffman table 2, and the Huffman code index value L〇〇kup[m,] of the lookup table 30 is Huffman code index value n' (n, = LG 〇 kup [m,]). Therefore, according to the leader of the bit string to be decoded! The number of the reference index m can be obtained for the corresponding lookup table. Thereafter, based on the lookup index value m, the corresponding Huffman code index value Looki^m,] can obtain the Huffman code index η' of the Huffman table 2(). Finally, according to the Huffman code index value n, the symbol, the bit length and the Huffman code are specified. In order to make the decoding method of the traditional preamble 1 search more clear and easy to understand, the lower left will take the bit string 〇 1] 1011 as an example: first, the number of poems of the bit rate 0111011 is equal to 〇, that is, the view The reference index value m of Table 3G is then obtained by referring to the index value m, which is equal to 0, and the value of the lookup table 3 又 = code index value Lookup [m, 卜 L 〇〇 kup [〇] = 〇. Followed by the Huffman code index value LGGKup_ by Chapi 2, the Huffman table 2〇门二^引值11' is equal to G, and from the Huffman code index value Π, equal to (start search Brother Hoffman Table 20. 1343192

, it TW5226PA Because the Hoop code index value n is equal to the Huffman code corresponding to 〇 and the bit length is 2, so the bit string 〇 11〇u the first two most important bits 01 and Hoffman Pallet comparison. Since 〇 1 does not match 〇〇, then the Huffman table 20 is searched down. The next Huffman code index value η is equal to 丨. Since the Huffman code index, the value η' is equal to 1 and the Huffman code is 010 and the bit length is 3, the first three most significant bits 位11 of the bit string 0111011 are compared with the Huffman code 010. Since 011 and 010 do not match, the Huffman table is then searched down. • A Huffman code index value η' at 20 ° is equal to 2. Since the Huffman code index value η' is equal to 2 and the Huffman code is 011 and the bit length is 3, the first three most significant bits 位11 of the bit string 0111011 are compared with the Huffman code 〇11. Since 011 and 011 match, the search for the Huffman table 20 is stopped. When the Huffman code index value η' is equal to 2, Sym[n,] = Sym[2] = 2, the symbol of the solution is 2 and the bit length is equal to 3. Finally, the three most significant bits in the bit string 0111011 are removed to produce the removed bit string 1〇11. In order to make the decoding method of the traditional preamble 1 search more clear and easy to understand, the following takes the bit string 0010 as an example: first, the number of bits before the bit string 〇〇1〇 is equal to 0, that is, the lookup table 30 refers to the index value m, equal to 〇. Then, by referring to the index value m' equal to 〇, the Huffman code index value Lookup[m']=Lookup[0]=〇 of the lookup table 3〇 is known. The Huffman code index value η' of the Huffman table 2〇 is equal to 0, and the Huffman code index value n is obtained by looking up the Huffman code index value Lookup[0]=0 of Table 3. 'Equivalent to 〇 start ^ find Hoffman table 20. Since the Huffman code index value η' is equal to 〇, the Huffman code is 〇〇 5 -Ϊ343192

TW5226PA and the length of the unit is 2, so the first two most significant bits 00 of the bit string 〇〇1p are compared with the Huffman code 00. Since 00 and 00 match, the search for the Huffman table 20 is stopped. When the Huffman code index value n is equal to 〇:

Symtn'hSymtO] = 0', the symbol of its solution is 〇 and the bit length is equal to 2. Finally, the 2 most significant bits in the bit string 〇_ are removed to generate the shifted bit string 10. It can be seen that when the conventional preamble 1 search method searches for the bit string non1 non and the bit string ooio, at least 4 times need to be searched for the corresponding symbol 2 and symbol 0. & beta month refers to Figure 4, which is a look-up table for the traditional One-bit Binary Search. Look at Table 4, including the Block Index and the Field Content. The field Index includes the node Node and the bit Bit' and the block content includes the indication value NodeSE and the flag Flag. According to the node Node and the bit Bit, it will correspond to different indication values NodeSE and flag Flag. Note that the Node value at the start of the search must be 0, and the Bit value is entered one by one according to the bit string. When the flag Flag is equal to 1, ' indicates that the binary search has ended, and the indication value NodeSE is the symbol corresponding to this bit string. Conversely, when the flag Flag is equal to 0, it indicates that the binary search has not ended, and the indication value NodeSE indicates the next node Node. Another binary search is performed according to the next node Node and the next bit Bit. By searching the lookup table 40, the symbol corresponding to the received bit string can be found. In order to make the decoding method of the traditional one-bit binary search more clear and easy to understand, the following will take the bit string 〇 111011 as an example: First, the node Node is equal, and the first bit of the bit string is 1343192

' t TW5226PA

Bit is equal to "0". The first binary search is performed according to the node and the bit element, and the flag value Flag and the indication value NodeSE corresponding to the node and the bit element are found. At this time, the flag value Flag is "0", and the indication value NodeSE is "1". Since the flag value Flag indicates that the search has not ended, and the indication value NodeSE is ''1', indicating that the next node is "" 1 ''. ^ Next, the node Node is equal to "1", and the second bit of the bit string is equal to "1". The second binary search is performed according to the node "1" and the bit "1" to find out The flag value Flag corresponding to the node "1" and the bit "1" and the indication value LuNodeSE. At this time, the flag value Flag is "0"' and the indication value NodeSE is ''3". Flag is ''0', indicating that the search has not ended yet, and the indication value NodeSE is "3", indicating that the next node Node is "3". Then, the node Node is equal to "3"', and the third of the bit string The bit is equal to "1". The third binary search is performed according to the node "3" and the bit "1", and the flag value Flag corresponding to the node "3" and the bit "1" is found and the indication value NodeSE. At this time, the flag value Flag is "1", and the indication value NodeSE is ''2''. Since the flag value Flag is "1", indicating the end of the search, and indicating that the φ value NodeSE is ,, 2, ', the symbol of the solution is ,, 2,,. For the decoding method of the conventional one-bit binary search. More clear and easy to understand, the following will be further illustrated by the bit string 0010 as an example: First, the node Node is equal to and the first bit of the bit string Bit is equal to "0". According to the node "0" and the bit "0" The first binary search is performed to find the flag value Flag and the indication value NodeSE corresponding to the node and the bit. At this time, the flag value Flag is "0", and the indication value NodeSE is "1". The flag Flag indicates that the search has not ended, and the indication value NodeSE is ''1'', indicating that the next node Node is ''1''. 7 inches 343192

TW5226PA Next, the node Node is equal to "1", and the second bit of the bit string * » is equal to "0". The second binary search is performed according to the node "1" and the bit "0", and the flag value Flag and the indication value NodeSE corresponding to the node ''1" and the bit are found. At this time, the flag value Flag is " 1", and the indication value NodeSE is. Since the flag value Flag is ''1', it indicates that the search ends. The indicator value NodeSE is a symbol indicating that the solution is solved. It can be seen that the conventional one-bit binary search method searches for the bit string 0111011 and the bit string 0010, and at least 5 times need to search for the corresponding symbol 2 and symbol 0. SUMMARY OF THE INVENTION According to this embodiment, a decoding method is proposed. The decoding method includes: obtaining the mth lookup index value of the newly created lookup table according to the number of leading bits of the bit string (Number of Leading One's), and the mth lookup index value is equal to the number of the leading string of the bit string; Obtaining the nth reference Huffman code index value of the reduced Huffman table according to the mth lookup index value, and the nth reference Huffman code index value is used to index the nth reference Hough of the reduced Huffman table Mann code and nth reference symbol index value, and the reference Huffman code index value is equal to the bit length of the nth reference Huffman code minus 1; the bit string is obtained according to the nth reference Huffman code index value η+1 Most Significant Bit (MSB); generate a difference value according to n+1 most important bits and the nth reference Huffman code; determine whether the difference value is less than zero; if the difference value is not If it is less than zero, a symbol index value is generated according to the difference value and the nth reference symbol index value; the symbol is obtained according to the symbol index value. If the difference value is less than zero, then decrease or increment the value of η and continue back to the streamlined Huffman table. 1343192

-TW5226PA Continue to compare difference values. 4 4 This paper is only described for Huffman tables applied to JPEG, but this embodiment can cover Huffman tables for other applications. For example, the longest Huffman code of the JPEG standard is 16 bits, so the maximum number of leading ones must be less than 17. In other applications, these numbers are not necessarily 16 and 17, respectively. In order to make the above-mentioned contents of the present embodiment more obvious, the following embodiments are specifically described, and the detailed description is as follows: [Embodiment] In order to reduce the complete Huffman table and look-up table Table size and increase decoding speed and table reconstruction speed. The description of the first embodiment and the second embodiment will be exemplified below. First Embodiment Please refer to Figs. 2 and 5 simultaneously, and Fig. 5 shows a simplified Huffman table which is the first embodiment. The first embodiment will provide a streamlined Huffman table 50 which, in order to distinguish it from the complete Huffman table 20, will be referred to as a reduced Huffman table 50. The reduced Huffman table 50 records only the partial Huffman codes and Huffman code index values in the complete Huffman table 20, and the Huffman codes recorded in the Huffman table 50 are hereinafter referred to as benchmarks. The Huffman code HfmBase[n].lstCode calls it, and the symbol index value recorded by the reduced Huffman table 50 is referred to below as the reference symbol index value HfmBase[n].lstOfs. Streamlined Huffman Table 50 includes reference symbol index value field and benchmark Huo 9 1343192

TW5226PA Fuman code field, not _. base transfuscent value n in the reference symbol index field corresponds to different reference symbol index value *

HfmBaSe[n]. 1 st〇fs, and different reference Huffman code index values η correspond to different reference Huffman codes in the reference Huffman code clamp

HfmBase[n].lstCode. Please note: when the reference Huffman code index value (n) in Figure 5 is equal to 〇, 9, 10, 11, 12, 13' 14, 15, the corresponding reference symbol index value and the reference Huffman code It is invalid, because when n is equal to 〇, 9, 1〇, u, 12, 13, 14, 15, its >1(:[11] is equal to zero', which means that the length is 11+1. The Fuman code does not exist at all. That is, there is no Huffman code of length 〖, 1〇, u, 12, 13, 14, 15, 16 bits. Unlike the complete Huffman table record all Hoffman Code, the reduced Huffman table 50 is the same bit length, only the Huffman code with the smallest code value is recorded as the reference Huffman code HfmBase[n].lstCode, and the reference Huffman code HfmBase[n is recorded. The reference symbol index value HfmBase[n].lstOfs corresponding to .lstCode. In addition, the reference Huffman code index value n is equal to the bit length of the corresponding reference Huffman code HfmBase[n].lstCode minus 1. For example The Huffman code of the complete Huffman table 20 with a bit length of 3 includes 010, 0U, 1〇〇, 1〇1, and 11〇. The reduced Huffman table 5〇 does not record 010, 011, 1 at the same time. 〇0, 101 and 11() Huffman code with a bit length of 3, but only the smallest code value is recorded as the reference Huffman code HfmBase[2].lstCode, and the index value corresponding to the Huffman code 〇10 is recorded. 1 as the reference symbol index value HfmBase[2] lst〇fs. The reference Huffman code 010 corresponding to the reference Huffman code index value n is equal to the bit length of the reference Huffman code 010 minus 1 ^ So, the reference Hoff曼码〇1〇 corresponding base 1343192

r TW5226PA The quasi-Huffman code index value n is equal to 2. • * The nth reference Huffman code HfmBase[n].lstCode of the Huffman table 50 and the nth reference symbol index value HfmBase[n].lstOfs are based on the symbol array Sym□, including each bit length The Huffman code number array NC[] and its index value η are generated. The nth reference Huffman code HfmBase[n]. IstCode is equal to the Huffman code of the original Huffman table, where all bit lengths in 20 are (n+1) and the code value is the smallest. For example, the Huffman code having a bit length equal to 3 in Fig. 2 includes 010, φ 0H, 100, 101, and 110. Since the code value of the Huffman code 010 among the Huffman codes 010, 0H, 100, 101, and 110 is the smallest, in FIG. 5, the second reference Hough of the Huffman table 50 is simplified. Man code HfmBase[2].lstCode=010. In addition, the nth reference symbol index value of the Huffman table 50 is reduced.

HfmBase[n]. 1 stOfs corresponds to the nth reference Huffman code HfmBase[n].lstCode. The nth reference symbol index value HfmBase[n]. 1 stOfs is equal to the index value of the symbol represented by the nth reference Huffman code HfmBase[n].IstCode in the symbol array Sym□. For example, the symbol corresponding to the Huffman code 010 in Fig. 2 is 1. Since the index value of the symbol 1 in the symbol array Sym□ is also 1, the second reference symbol index value HfmBase[2].lst0fs=l in the reduced Huffman table of Fig. 5. In order to make the present embodiment more clear and understandable, a method of generating a reduced Huffman table 50 will be exemplified below, but this example is not intended to limit the scope of the patent application. Insulation 343192 TW5226PA, , please refer to Fig. 6, which is a flow chart for generating the reduced Houghman table of the first embodiment. First, as shown in step 410, the 0th reference symbol index value HfmBase[0].lst〇fs of the reduced Huffman table 50 is set equal to 0, and the 0th reference Huffman code of the reduced Huffman table 50 is set. HfmBase[0], lstCode is equal to 0. Next, as shown in step 420, the reference Huffman code index value η is set equal to one. Next, as shown in step 430, it is determined whether the reference Huffman code index value η is less than 16 ? If not, the flow is ended. If yes, step 440 is performed. Step 440: generating an nth reference symbol index value HfmBase[n].lstOfs according to the nl ❿ reference symbol index value HfmBase[nl].lstOfs and the number of Huffman codes NC[nl] of length η Η_1 reference Huffman codes HfmBase[n· 1 ]. 1 stc〇de and the number of Huffman codes of length η NC[n-1 ] Generate the nth reference Huffman code HfmBase[n].lstCode . Wherein the nth reference symbol index value HfmBase[n].lstOfs =HfmBase[nl].lstOfs + NC[nl] ' and the nth reference Huffman code HfmBase[n].lstCode, etc. at the η·1 The sum of the reference Huffman code HfmBase[nl].lstCode and the number of Huffman codes of length η, NC[nl], is multiplied by two. The reference Huffman code index value η is then incremented as in step 450. If the incremented reference Huffman code index value η is still less than 16 ', step 44 and step 45 are repeated. Wherein, the number of Huffman codes whose length is η\(:[11_1] is obtained by an array of Huffman codes of the same bit length Nc[]. The generation of the reduced Huffman table 50 can be varied. Program writing state 12 1343192

*, TW5226PA-like, for illustrative purposes, the following will give an example, but this example is not limited to the scope of the patent application. The streamlined Huffman table 50 is implemented, for example, by the following procedures:

HfmBase[0].lstOfs = 0;

HfmBase[0].lstCode = 0; for(n =1; η <16; η++) {

HfmBase[n].lstOfs = HfmBase[n-l].lstOfs + NC[n-l]; 肇 //1 st offset of same length codes

HfmBase[n].lstCode = (HfmBase[nl].lstCode + NC[nl]) « 1; // 1 st code of same length codes Please refer to Figure 5 and Figure 7 together, and create a new lookup table 7 The reference Huffman code index value Lookup[m] is established based on the index value n of the Huffman code HfmBase[n].lstCode of the reduced Huffman table 50. It should be noted that 'm is equal to the leading number of the Huffman code, and the reference Huffman code index value Lookup[m] is equal to the reference Huffman code HfmBase with m leading 1 in the reduced Huffman table 50 [ n]. 1 stCode also indexes the value η. Therefore, Lookup[m] = η, which indicates that the number of leading ones of the nth HfmBase[n].lstCode in the reference Huffman code of the reduced Huffman table 50 is equal to m. For example, when the index value m of the new lookup table 70 is equal to 4,

Lookup[m] = Lookup[4] = 4 = η, which means the first 13 1343192 TW5226PA of the 4th reference Huffman code HfmBase[4].lstCode(= 11110) of the Huffman table 50 Equal to 4. 4 However, there may be two exceptions to the above description. The exception is that there are more than two basal quasi-Hochiffs in the Huffman Table 50. The number of leading rrrrrustc°de is the same. Since the plurality of reference Huffman codes HfmBase[n]"stcode of the same number and the same number are respectively 1 to the base Huffman code index value n, at this time, the mth reference of the newly created lookup table 70 The husbandman code index value selection is equal to these benchmarks: the largest of the index values n. For example, streamlined Huffman's stagnation benchmark Huffman code HfmBase[1] lstc〇de and the second benchmark Hoffman mom HfmBase[2].丨StCode's leading number of condensed Huffman table 5 The second base Huffman code index value 2 of 〇 is greater than the first, quasi-Huffman code index value! Therefore, the new benchmark Huffman code for the new lookup table 7() is Lookup[〇]=2. The second exception is the lookup index value of the new lookup table 7 m method = the reference Huffman code HfmBase[n] 精 of the reduced Huffman table 5〇, then the mth reference Huffman code index The value L〇〇kup[m] is equal to the m·1th base Huffman code index value Lookup[ml]. For example, the baseline Huffman code index values L〇〇kup[l] and Lo〇kup[2] of the new lookup table 70 cannot be obtained from the reduced Huffman table 50, that is, the reduced Huffman table 50 The number of leading i of the reference π having a reference Huffman code HfmBase[n]"stC〇de is equal to 1 or 2. At this point, let the reference Huffman code of the newly created lookup table 7〇 Lo〇kup[2]=Lookup[l]=Lookup[0]=2. Please refer to Fig. 7, which shows a newly created look-up table which is a simplified Huffman table of the first embodiment. In order to distinguish it from the traditional look-up table, the following reduced Huffman table lookup table will be referred to as the new lookup table 70. Streamlined Huffman Table 5〇 1343192

', TW5226PA • The new lookup table 70 includes the reference Huoman $man code index value position, the reference Huffman code index value (4) to record each reference (four) value m corresponding to the base Huffman code index value L〇〇kuP [m] 'The reference Huffman code index value L〇〇kuP[m] is the reference Huffman code index value η of the reduced Huffman table 5〇. Its towel's time-valued m is equal to the number of bits before the bit string to be decoded. In order to make the present embodiment clearer (4), the following will exemplify a method of generating a new lookup table 7G', but this example is not within the scope of the patent application. Please note that when the reference index value (m) is greater than 8 at the 7th, the reference value of the reference Huffman code corresponding to Lu is invalid, because the maximum valid reference Huffman code index value (8) in the $Fig is 8 , corresponding to Nc[n] is 1 ' and its corresponding HfmBase[n].istCode* U11(1)丨〇, that is, the maximum code of this Huffman table is 11U 1U1 〇. Therefore, the maximum number of leading ones is 8, that is, the maximum index value m is 8. Referring to Figure 8, there is shown a flow chart for creating a new lookup table of the reduced Houghman table of the first embodiment. First, as shown in step 612, the reference index value m is equal to the reference Huffman code index value η is equal to, as shown in step 614, determining whether the reference Huffman code index value η is less than 16 9 if the reference Huffman code index value „Not less than 16, step 63〇 is performed. Conversely, if the reference Huffman code index value η is less than 16, then step 616 is performed. Step 616 is to calculate the nth reference Huffman code.

HfmBase[n]. 1 stCode leads to 1 number l. Since the condensed Huffman table 50 does not record all of the Huffman codes, the partial Huffman code index value L00kup[m] of the new _ 7〇 will not be directly known by the condensed Huffman table 50. For example, a Huffman code with a bit length of 3 in the complete Huffman table 20 includes a look at 1343192

TW5226PA index value m' is equal to 0. 010 and Oil, the reference index value m' is equal to 1 of 100 and 101, and the reference index value m' is equal to 2 of 110, so the lookup table 30 can be directly obtained from the complete Huffman table 20. It is known that the index value m' is equal to the Huffman code index value Lookup[m'] is equal to 0, and the index value m is consulted, and the Huffman code index value Lookup[m'] equal to J is equal to 3 and the index value m is consulted. The Huffman code index value Lookup[m'] equal to 2 is equal to 5. However, the reduced Huffman table 50 only records m equal to 0 010. Therefore, when the reference value m is equal to 1 or 2, the partial reference Huffman 蝎 index value Lookup[m] of the new lookup table 70 will It cannot be directly learned by the streamlined Huffman Table 50. Therefore, step 618, step 620 and step 622 are executed to fill the partial reference Huffman code index value Lookup[m] which cannot be directly known by the reduced Huffman table 50, such as the reference Huffman code index value Lookup [l] ] and the base Huffman code index value Lookup [2]. Step 618 is to determine whether the reference index value m is smaller than the nth reference Huffman code HfmBase[n]. 1 stCode is preceded by a number L? If the reference index value m is not smaller than the nth reference Huffman code HfmBase[n ].lstCode is preceded by a number L, which means that m is equal to L, since m cannot be greater than L, that is, there is no missing leading number, and step 624 is performed. Conversely, if the reference index value m is smaller than the n-th reference Huffman code HfmBase[n].lstCode and the number L is preceded, it indicates that there is a missing leading number, and step 620 is executed. Step 620 is such that the mth reference Huffman code index value Lookup[m] in the new lookup table 70 is compensated by the mth-th order reference Huffman code index value L〇〇kup[m-l]. The index value m is then incrementally reviewed as indicated by step 622. Following step 618, step 620 and step 622 will be repeated until the reference index value m is not less than the nth reference Hough 1343192

^ TW5226PA Man code HfmBaSe[n].lstC〇de leads a number L before changing step 624. Step 624 is such that the Lth reference Huffman code index value L〇〇kuP[L] in the new lookup table 7 is equal to the reference Huffman code index value n. Then, as shown in step 026, the index value m is equal to the number of leading points of the eleventh reference Huffman code / HfmBaSe[n].lstC〇de [plus! It represents the number of the leading ones that should be next, that is, the index of the new creation of the newly created look-up table. Following the step 628, the increment is returned to step 614. '

Step 630, step 632 and step 634 are used to fill the corresponding reference Huffman code index value L〇Gkup[m] of the insufficient index index m, since either the complete Huffman table or the simplified Huffman The table can be recalculated = all 17 (〇 to 16) leading values. Step (10) is to judge whether the index value m is less than 17? If the reference index value is not less than 17, the end of the flow plan number 1 is not expected to be 17. Conversely, if the index value m is less than 17, then step (4) is performed. = read the table "recording Su ^ ^ LGGku] J and so on = the second reference Huffman code (four) value L (8) _ (four) is also a filling function as step 620. Index value m, and return to step 63. The incremental review shown in 634 produces a new lookup table. At, 描, ^ 乂 乂 夕 夕 夕 = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = 1343192

TW5226PA for(m = η = 0; η <16; η++) { // for all code lengths·... 4 L = leading_one(HfmBase[n].lstCode); // get number of leading 1 's from Simplified table for(; m <L; m++) // pad the gap in Lookup[ ] if there is

Lookup[m] = Lookup[m - 1];

Lookup [L] = n; // save (length - 1) to Lookup[ ]. Overwrite is possible, m = L + 1; } for(; m <17; m++) // pad the trailing elements in Lookup[ ]

Lookup[m] = Lookup[m - 1]; The function leading_one() in the program above is used to calculate the number of leading 1's of the bit string. Referring to Figure 9, there is shown a flow chart of the decoding method in accordance with the first embodiment. First, as shown in step 710, the lookup index m of the newly created lookup table 70 is obtained according to the number of leading bits of the bit string to be decoded, and the index value m is used to indicate that the bit string is preceded by one. Since the reference index value m is used to indicate the number of bits before the bit string is forwarded, when the reference index value m is equal to 1, it indicates that the number of leading bits of the bit string is equal to 1; when the index value m is equal to 2, The number of bits before the bit string is equal to 2, and so on. In other words, the reference index value m is equal to the bit string leading to 1343192

'TW5226PA number. # Next, as shown in step 720, the reference Huffman code index value η of the reduced Huffman table 50 is obtained from the lookup index value m, and the reference Huffman code index value η is used to simplify the Huffman table 50. The nth reference Huffman code HfmBase[n].lstCode and the nth reference symbol index value HfmBase[n]. 1 stOfs, and the reference Huffman code index value 11 is equal to the nth reference Huffman code HfmBase The length of the bit of [n].lstCode is reduced, wherein the reference Huffman code index value η is equal to the reference Huffman code index value • Lookup[m]. Next, as shown in step 730, n+1 of the most significant bits of the bit string are taken from the reference Huffman code index value η. The n+1 most significant bits of the 'bit string' are indicated by ν in the first embodiment. Since the reference Huffman code index value η is equal to the bit length of the nth reference Huffman code HfmBase[n].lstCode minus 1, the first Huffman code index value η can be obtained in step 730. The bit length of n reference Huffman codes HfmBase[n].lstCode. Then, according to the bit length of the nth reference Huffman code φ HfmBase[n].lstCode, the n+1 most important bits of the bit string are obtained. Then, as shown in step 740, a difference value Diff is generated from the n+1th most significant bit ν and the nth reference Huffman code HfmBase[n].lstCode. Wherein, the difference value Diff is equal to η+1 most important bits ν minus the ηth basis Huffman code HfmBase[n].lstCode. Then, as shown in step 750, is it judged whether the difference value Diff is less than zero? If the difference value Diff is not less than zero, as shown in step 760, the symbol 1343192 is generated according to the difference value Diff and the nth reference symbol index value HfmBase[n].lstOfs.

TW5226PA index value Ofs. The symbol index value Ofs is, for example, equal to the difference value Diff plus the η*·th reference symbol index value HfmBase[n].lstOfs. Next, as shown in step 770, the symbol, i.e., S, is obtained from the symbol index value Ofs. Wherein, step 770 obtains a corresponding symbol from the symbol array Sym□ in the header according to the symbol index value Ofs, that is, S is equal to Sym[0fs]. Following the step 772, the n+1 most significant bits of the bit string are removed to produce a removed bit string. Conversely, if the difference value Diff is less than zero, as shown in step 780, the reference Huffman code index value η is adjusted in a decreasing manner and then returns to step 730. In order to make the decoding method of the first embodiment more clear and easy to understand, the following description will be made by taking the bit string 0111011 as an example: first, as shown in the foregoing step 710, the number of leading bits 1 of the bit string 0111011 is equal to 0. The lookup index value m of the newly created lookup table 70 is equal to 0 according to the number of the first bit of the bit string 0111011. Next, as shown in the previous step 720, the reference Huffman code index value η of the reduced Huffman table 50 is equal to the reference Huffman code index value Lookup[0] of the new lookup table 70 record is equal to two. Following the previous step 730, the three most significant bits 011 of the bit_ are taken according to the reference Huffman code index value η equal to two. Then, as shown in the foregoing step 740, the second reference Huffman code HfmBase[2].lstCode of the reduced Huffman table 50 and the second reference symbol index are searched for the first time according to the reference Huffman code index value η. The value HfmBase[2].lst0fs. The second base Huffman code HfmBase[2].lstCode of the reduced Houghman table 50 is 010, and the second reference symbol index value of the reduced Huffman table 50 is 1343192.

TW5226PA

HfmBaSe[2], 1 St〇fs is 丨. The acoustic value Diff is equal to the three most significant bits 011 of the bit string minus the second reference Huffman code 010 equal to i. Then, as shown in the foregoing step 750, is it judged whether the difference value Diff is less than zero? Since Diff is equal to 丨', the difference value Diff is not less than zero. Therefore, as shown in the foregoing step 760, the symbol index value 〇fs is equal to the difference value Diff plus the second reference symbol index value HfmBase[n]"st〇fs. Where the difference value Diff is equal to 1 and the second reference symbol index value is eight

HfmBaSe[n].1 St〇fs is equal to Bu, so the symbol index value Ofs is equal to 2. Next, as shown in the foregoing step 770, the symbol s special Sym [[fs]] is obtained from the symbol array sym[] in the header according to the symbol index value Ofs. For example, the symbol array is, for example, (10) two from two to ice (1). Therefore, when the symbol index value Ofs is equal to 2, the obtained symbol is equal to _(7) equal to 2. Following the previous step 772, the three most significant bits oil of the bit string are removed and the removed bit string 1〇11 is obtained. In order to make the above decoding method more clear and easy to understand, the following will take the example of bit 0010 as an example: firstly, as shown in step 710, the number of the other derivative of the bit string 00H) is equal to G, according to the bit

The number of conversions to the newly created lookup table is checked. (4) The value m is equal to G. The description of step 720 is shown in step 720. The reference code of the Huffman table 50 is equivalent to the value of L. Π is equal to the benchmark of the new record. Hoffman according to the benchmark Huo: shown, the root _. ‘, equal to 2 to get the three most important bits of the string. The second time of the search for the simplified Hoffman table, the second reference Hoffmanma 1343192

TW5226PA

HfmBase[2].lstCode and the second reference symbol index value * ·

HfmBase[2].lstOfs. The second reference Huffman code HfmBase[2].lstCode of the reduced Huffman table 50 is 010, and the second reference symbol index value HfmBase[2].lstOfs of the reduced Huffman table 50 is 1. The difference value Diff is equal to the 3 most significant bits 001 of the bit string minus the second reference Huffman code 010 equals -1. Then, as shown in the foregoing step 750, is it judged whether the difference value Diff is less than zero? Since Diff is equal to -1, the difference value Diff is less than zero. Therefore, the reference Huffman code index value η will be decremented as shown in the previous step 780. The decremented reference Huffman code index value η is changed from 2 to 丨. Step 730 is repeated to obtain the two most significant bits 00 of the bit string according to the reference Huffman code index value η equal to one. Step 740 is repeatedly executed to search for the first reference Huffman code HfmBase[l].lstCode of the reduced Huffman table 50 and the first reference symbol index value according to the reference Huffman code index value η equal to 1 for the second time. HfmBase[l].lstOfs. The first reference Huffman code HfmBase[l].lstCode of the reduced Huffman table 50 is 00, and the first reference symbol index value HfmBase[l].lstOfs of the reduced Huffman table 50 is 0. The difference value Diff is equal to the 2 most significant bits 00 of the bit string minus the 1st reference Huffman code 00 is equal to 0. Repeat step 750 to determine if the difference value Diff is less than zero? Since Diff is equal to 0, the difference value Diff is not less than zero. Therefore, as shown in the foregoing step 760, the addition difference value Diff and the first reference symbol index value HfmBase[l].lstOfs generate the symbol index value Ofs. Since the difference value Diff is equal to 0 and the first reference symbol index value HfmBase[l].lst〇fs 22 1343192

· · TW5226PA IT binary symbol index value 0 color is equal to. . Then execute the step ". = index value ofs from the symbol array in the header - the acquisition 5 tiger S is equal to Sym [0fs]. For example, the symbol array is for example {〇'1,2,3,4,5,6 , 7, 8, 9, 10, 11}. Therefore, when the symbol index value (10) is equal to 0, the 'get symbol is equal to Sym[0] is equal to 〇. Following step 772, the 2 most significant bits of the bit string 0010 are removed. The element 'is obtained after the removed bit string 10. The above decoding method makes the bit string 011011 and 0010 only need to compare 3 times and can quickly obtain the corresponding symbol 2 and symbol 〇. For the second embodiment, please refer to the second FIG. 10 and FIG. 10 show that the nth reference Huffman code HfmBase[n] lstC〇de and the second reference symbol index value HfmBaSe[n].ist0fs of the Huffman table 100 are based on the symbol array Sym[], The Huffman code array Nc□ and the index value η are included for each bit length. The nth reference Huffman code HfmBase[n] lstC〇de is equal to all bits in the original Huffman table 20. The Huffman code whose length is (n+1) and whose code value is the largest. For example, the Huffman code whose bit length is equal to 3 in FIG. 2 includes 010, 011, 1〇〇. , 101 and 11. Since the code value (c〇de Va丨ue) of the Huffman code 110 among Huffman codes 〇11, 100, 101 and 110 is the largest, in Figure 10, the simplification is simplified. The second reference Huffman code HfmBase[2].lstCode=110 of the Huffman table 1 is further reduced. In addition, the nth reference symbol index value HfmBase[n].lstOfs of the Huffman table 100 is corresponding to the η reference Huffman codes HfmBase[n].lstCode. nth reference symbol index value 23 1343192

TW5226PA « *

HfmBase[n].lst〇fs, etc., nth reference Huffman code HfmBase[n]. 1 The index of the symbol represented by stCode in the symbol array Sym[]. For example, the symbol corresponding to the Huffman code 110 in Fig. 2 is 5. Since the index value of the symbol 5 in the symbol array Sym[] is also 5, the second reference symbol index value HfmBase[2].lstOfs=5 in the reduced Huffman table 100. Referring to Figure 10, there is shown a reduced Huffman table which is a second embodiment. In the foregoing first embodiment, the Huffman code with the smallest code value and the corresponding index value are selected as the reference Huffman code HfmBase[n]. 1 stCode and in the Huffman codes of the same bit length. The reference symbol index value HfmBase[n].lstOfs. The reduced Huffman table (10) of the second embodiment is different from the reduced Huffman table 50 of the first embodiment in that the second embodiment selects code values in Huffman codes of a plurality of identical bit lengths. The largest Huffman code and its corresponding index value are used as the reference Huffman code HfmBase[n]. I stCode and the reference symbol index value HfmBase[n].lstOfs. For example, a full Huffman table 1 霍 Holmanman code with a bit length of 3 includes 010, 011, 100, 101, and 11 〇. The simplified Huffman table 100 is not the Huffman code 010 that records the smallest code value in 010, 011, 1〇〇, 1〇1, and 丨1〇, but only the 110th code value is the base Huffman code. HfmBase[2].lstCode, and record the index value 5 corresponding to the Huffman code 110 as the reference symbol index value HfmBase[2].lst0fs. The reference Huffman code index value η corresponding to the reference Huffman code 110 is equal to the bit length of the reference Huffman code 〇 1 减 minus one. Therefore, the reference Huffman code index η corresponding to the reference Huffman code 11 等于 is equal to two. 24 1343192

• * TW5226PA Please, meaning: when the reference Huffman code in Figure 10, the reference (n) is equal to 〇, 9' 1 〇 ' 11 ' 12, 13, 14, 1 5, the corresponding reference symbol Both the index value and the reference Huffman code are invalid, because when η is equal to 〇, 9, 10, 11, 12, 13, 14, 15 'its NC[n] are equal to zero, which means that the length is ^ + The Huffman code of 1 does not exist at all. Please refer to Fig. 10 and Fig. 11 at the same time. The reference Huffman code index value Lookup[m] of the newly created lookup table is based on the Huffman code HfmBase[n] of the simplified Huffman table. The index value of lstCode is established by n. # It should be noted that m is equal to the number of leading i of the Huffman code, and the reference Huffman code index value Lookup[m] is equal to the base of the reduced Huffman table 1 with m rigid guides 1 The index value η of the Fuman code HfmBase[n] 1 stC〇de. Therefore, L〇〇kuP[m]=n indicates that in the reference Huffman code of the reduced Huffman table, the number of leading ones of the nth HfmBase[n].lstC〇de is equal to m. For example, when the index value of the new lookup table 11 is equal to *, its L〇〇kuP[m] = L0〇kuP[4] = 4 = n, which means the fourth of the simplified Huffman table 100 The base Huffman code HfmBase[4].lstCode(= lliio) • The number of leading 1s is equal to 4. However, there are two exceptions to this description. The first exception is to have more than two benchmark Huffman codes in the Huffman table 100. HfmBase[n]. The number is the same. Since the plurality of reference Huffman codes HfmBase[n]"stc〇de having the same number of the first ^1 numbers respectively correspond to the plurality of reference Huffman code index values η, the mth reference of the new lookup table 11 The Huffman code index value is equal to the smallest of these reference Huffman code index values η. Since there is no first exception in Figure 1, it cannot be illustrated in Figure 10. However, suppose a streamlined Huo 25 1343192

The first base Huffman code HfmBase[l].lstCode of the TW5226PA Fuman table and the leading number of the second reference Huffman code HfmBase[2].lstCode are both equal to zero. Since the first reference Huffman code index value 1 of the reduced Huffman table is smaller than the second reference Huffman code index value 2, the third of the newly created lookup tables corresponding to the reduced Huffman table The reference Huffman code index value Lookup[0]=l. The second exception is the new lookup table. The reference index value m of 1〇 cannot be obtained from the reference Huffman code HfmBase[n].lstCode of the reduced Huffman table 100, then the mth reference Huffman code index The value L 〇〇 kUp [m] is equal to the m+1th reference Huffman code index value Lm > kUp [m + l]. For example, the baseline Huffman code index value LookupH of the new lookup table 11〇 cannot be obtained by the reduced Huffman table 1也. = There is no reference Huffman code in the reduced Huffman table 100.

The leader of HfmBase[n].lstCode! The number is equal to the value of the reference Huffman code index L〇〇kup[1] = Lookup[2] = 2 of the new lookup table 110. Please refer to Fig. 11 for a new look-up table which is a simplified Huffman table of the second embodiment. Since the condensed Huffman table of the second embodiment _ = the condensed Huffman table 50 of the first embodiment is different, the new lookup table 11 corresponding to the condensed Huffman table 100 is naturally corresponding to the condensed Hoffman Table 50 of the new creation checklist 7 is different. It should be specially stated that the 杳 杳 爪 〇 〇 〇 〇 〇 霍夫 霍夫 霍夫 霍夫 霍夫 H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H When the index value (10) is greater than 8, its 26 1343192

The reference Huffman ♦ code index value corresponding to TW5226PA is invalid because the maximum valid reference Huffman code index value (n) in the first diagram is 8, which corresponds to HfmBase[n]. lstCode is ini η〗 〖, that is, the maximum code of this Huffman table is 1111 1111 0. Therefore, the maximum number of leading ones is 8, that is, the maximum index value m is 8. Since the reference values selected in the first embodiment and the second embodiment are different, the contents of the simplified Huffman table and the newly created look-up table are different from each other. Although the production method is different, the spirit of the generation is one, so that the simplified Huffman table 100 and the new look-up table 110 of the second embodiment are not described herein. Referring to Fig. 12, it is shown that the decoding method according to the second embodiment is (4). First, as shown in the step chat, the reference index value m of the newly created lookup table 11 is obtained according to the number of bits before the string to be decoded, and the index value m is used to indicate the number of leading bits of the bit string. Since the reference index value m is used to indicate the number of leading digits before the bit string, when the index value m is equal to 】, the number of leading words before the bit string is equal to "丨• When referring to the index value m When it is equal to 2, it means that the number of leading bits (10) of the bit string is equal to 2' and so on. That is to say, the reference index value m is equal to one number before the bit string. Next, as shown in step 1020, the reference Huffman code index value η of the trick Huffman table 100 is obtained based on the lookup index value m, and the reference Huffman code ^ index η is used to index the reduced Huffman table. The nth reference Huffman drink HfmBaSe[n]. 1 stCode and the nth reference symbol index value

HfmBase[n].lst〇fs, and the reference Huffman code index value η is equal to the first hit Hoffman code HfmBase[n]. The bit length of i stCode is reduced! . Where 27 1343192 TW5226PA _ · The reference Huffman code index value η is equal to the reference Huffman code index value 4 *

Lookup[m] ° follows, as shown in step 1030, the n+1 most significant bits of the bit string are obtained from the reference Huffman code index value η. The n+1 most significant bits of the bit string are indicated by ν in the second embodiment. Since the reference Huffman code index value η is equal to the bit length of the nth reference Huffman code HfmBase[n].lstCode minus 1, the first Huffman code index value η can be obtained in step 1030. η reference Huffman codes

The bit length of HfmBase[n].lstCode. Then, the n+1 most significant bits of the bit string are obtained according to the bit length of the nth reference Luhoffman code HfmBase[n].lstCode. Then, as shown in step 040, a difference value Diff is generated according to n+1 most important bits ν and the nth reference Huffman code HfmBase[n].lstCode. Wherein, the difference value Diff is equal to the nth reference Huffman code HfmBase[n].lstCode minus n+1 most important bits ν. Then, as shown in step 1050, is it judged whether the difference value Diff is less than zero? If the difference value Diff is not less than zero, as shown in step 1060, the symbolic index value Ofs is equal to the nth reference symbol index value HfmBase[n].lstOfs minus the difference value Diff. Conversely, if the difference value Diff is less than zero, then as shown in step 1080, the reference Huffman code index value η is incremented and then returned to step 1030. Next, as shown in step 1070, the symbol, i.e., S, is obtained based on the symbol index value Ofs. Wherein, step 1070 obtains a corresponding symbol from the symbol array Sym[] in the header according to the symbol index value Ofs, that is, S is equal to Sym[0fs]. 28 1343192

• The TW5226PA then removes the bit bit as shown in step 1072 to generate a bit string after the removal. The most important thing for the state of the string is to make the DM grazing onion new - - a, η is a hundred first as in the previous step 1010, the number of the leader of the 70 string 0111011 is equal to 〇, according to the bit: = The number of 1011/preamble bits is 1 and the lookup index value of the newly created lookup table 110 is equal to 〇. Then, as shown in the foregoing step 1020, the index value n of the streamlined Hoffmann Huffman code is equal to the value of the new creation: 矿 矿 啊 啊 啊 〗 〗 〗 〗 〗 〇 〇 〇 〇 霍夫 霍夫 霍夫 又 又 又 又 又 又 霍夫 霍夫 霍夫 霍夫 霍夫 霍夫 霍夫[0] is equal! . Following the deletion of the preceding steps, the two most important bits 01 of the bit string are obtained from the reference Huffman code index value n equal to one. Then, as shown in the foregoing step 1040, the first "solid reference Huffman code HfmBase[l].lstCode and the first reference symbol of the simplified Huffman table 1" are searched according to the reference Huffman code index value η. Index value

HfmBaSe[1].lst0fs. The first reference symbol index HfmBase(1)〗 st〇fs of the Huffman Huffman Table 1〇〇 第 霍夫 H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H H Hey. The difference value is equal to the 2nd most significant bit of the 2nd reference Huffman code 00 minus bit string 01 is equal to -1. Then, as shown in the foregoing step 1050, is it judged whether the difference value Diff is less than zero? Since Diff is equal to -1, the difference value Diff is less than zero. Therefore, as shown in the previous step 1080, the reference Huffman code index value n is incremented. The incremented reference Huffman code index value η is changed from 1 to 2. Step 1030 is repeated, according to the reference Huffman code index value η is equal to 2 to obtain the bit string 29 1343192

TW5226PA 3 most important bits 〇 11. * Repeat step 1040 to search for the second base Huffman code of the reduced Huffman table 100 based on the reference Huffman code index value 11 equal to 2 for the second time.

HfmBase[2].lstCode and the second reference symbol index value

HfmBaSe[2].lst0fs. Streamlined Huffman Table 1 〇〇 2nd benchmark Huffman code chest 8 coffee [2].1 says 0 as 11〇, and streamlined Huffman table 1〇〇 2nd reference symbol index value HfmBase [2].lst〇fs is 5. Difference value

Diff is equal to the 2nd most important bit 011 of the 2nd reference Huffman code reduction bit string 011 is equal to 3. Repeat step 1〇5〇 to determine if the difference value Diff is less than zero? Since Diff is equal to 3, the difference value bet is not less than zero. Therefore, as shown in the above step 1060, the second reference symbol index value HfmBase[2], lst〇fs minus the difference value migration generates the symbol index value (10). Since the difference value Diff is equal to 3 and the second reference symbol index value HfmBaSe[2].lst〇fs is equal to 5 ', the symbol index value 〇fs is equal to 5_3 -2. Next, step 1〇7〇 is performed to obtain the symbol S equal to Sym[〇fs] according to the symbol index value 〇 pseudo symbol array Symbol. For example, the symbol array is, for example, 8 1 〇, 11}. Therefore, when the symbol index value 〇fs is equal to 2, the obtained symbol is equal to Sym[2] equal to 2. Steps 〇72 are followed to remove the three most significant bits ' of the bit string 〇111〇11 and the removed bit string 1 〇 1 J is obtained. In order to make the decoding method of the second embodiment clearer and easier to understand, the following description will be made by taking the bit string 0010 as an example: first, as in the foregoing step 1〇1〇, the bit 70 string 0010 is preceded by one bit. The number is equal to 〇, and the reference index value 1343192 of the newly created lookup table 110 is obtained according to the number of the leading bits before the bit string 0010.

• - TW5226PA m is equal to 0. • # Next, as shown in the previous step 1020, the reference Huffman code index value η of the reduced Huffman table 100 is equal to the reference Huffman code index value Lookup[0] of the new lookup table 110 record is equal to one. Following the previous step 1030, the two most significant bits 00 of the bit string are obtained based on the reference Huffman code index value η equal to one. Then, as shown in the foregoing step 1040, the first reference Huffman code φ HfmBase[l].lstCode and the first one of the reduced Huffman table 1 are searched for the first time according to the reference Huffman code index value η. Reference symbol index value

HfmBase[l].lstOfs. The first reference Huffman code HfmBase[l].lstCode of the reduced Huffman table 100 is 00, and the first reference symbol index value HfmBase[l].lstOfs of the reduced Huffman table 100 is 0. The difference value Diff is equal to the 1st reference Huffman code 00 minus the 2 most significant bits of the bit string 00 is equal to 0. Then, as shown in the foregoing step 1050, is it judged whether the difference value Diff is less than zero? Since Diff is equal to 0, the difference value Diff is not less than zero. • Next, as shown in the foregoing step 1060, the first reference symbol index value HfmBase[l].lstOfs minus the difference value Diff produces a symbol index value Ofs. Since the difference value Diff is equal to 0 and the first reference symbol index value HfmBase[l].lstOfs is equal to 0, the symbol index value Ofs is equal to 0-0 =0. Next, step 1070 is executed to obtain the symbol S equal to Sym[Ofs] from the symbol array Sym□ in the header according to the symbol index value Ofs. For example, the symbol array is, for example, {〇, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11}. Therefore, when the symbol index value Ofs is equal to 0, the obtained symbol is equal to Sym[0] equal to 0. Step 1072 is followed to remove the two most important 31 1343192 TW5226PA · bits of the bit string 0010, and the removed bit string 10 is obtained. • The Huffman decoding method disclosed in the above embodiment has a number of features. The larger the original Huffman table is, the more significant these features are: When the reduced Huffman table and its newly created lookup table are smaller than the complete Hoffman table and its traditional lookup table. Please refer to Figure 13, which is based on the Huffman table for AC brightness or AC chrominance provided by the JPEG standard. The drawing is based on the traditional lead 1 search (Leading 1's). Search), the traditional one-bit φ Binary Search and the table size comparison table of this example. As can be seen from Fig. 13, the table of the simplified Huffman table and its newly created look-up table of this embodiment is smaller than the complete Huffman table and its conventional look-up table. Please refer to Figure 14, which is also based on the Huffman table for AC brightness and AC chroma provided by the JPEG standard. The drawing is based on traditional lead 1 search, traditional one-bit binary search and this example. Comparison of the number of searches. Since the reduced Huffman table is smaller than the complete Huffman table, the reconstruction time of the reduced _ Huffman table and the new lookup table will be less than the reconstruction time of the complete Huffman table and the traditional lookup table, and the table is improved. Rebuild speed. In view of the above, although the above has been disclosed in various embodiments, it is not intended to limit the embodiment. It is to be understood that those skilled in the art can make various changes and modifications without departing from the spirit and scope of the embodiments. Therefore, the scope of protection of this embodiment is intended to be defined by the scope of the appended claims. 32 1343192

, TW5226PA [Simple description of the diagram] Figure 1 shows the Huffman table as the encoding end. Figure 2 depicts the Huffman table reconstructed for the complete decoder. Figure 3 depicts a look-up table for the Leading l’s Search. Figure 4 depicts a look-up table for a traditional One-bit Binary Search. Figure 5 is a simplified Huffman table of the first embodiment. Figure 6 is a flow chart showing the generation of the reduced Huffman table of the first embodiment. Fig. 7 is a diagram showing a new look-up table of the reduced Huffman table of the first embodiment. Figure 8 is a flow chart showing the creation of a new lookup table for the reduced Huffman table of the first embodiment. Fig. 9 is a flow chart showing the decoding method according to the first embodiment. Figure 10 is a simplified Huffman table of the second embodiment. Figure 11 is a diagram showing a new lookup table of the reduced Huffman table of the second embodiment. Fig. 12 is a flow chart showing the decoding method according to the second embodiment. Figure 13 shows the traditional Leading l's Search, the traditional One-bit Binary Search and the present application of the Huffman table applied to the AC brightness or AC chroma provided by the JPEG standard. A table size comparison table for the implementation example. 33 1343192

TW5226PA Figure 14 shows a comparison of the traditional pilot 1 search, traditional one-bit binary search and the search times of this example applied to JPE (Ji standard (AC standard brightness and AC chroma) Huffman table. Main component code 5^ δ 儿明] 10 : Huffman Table 20: Complete reconstruction Huffman table 50, 100: Streamlined Huffman table 70, 110: Newly created look-up table m', m: Lookup index value η ' : Huffman code index value η: reference Huffman code index values 410 , 420 , 430 , 440 , 450 , 612 ' 614 , 616 , 618 , 620 , 622 , 624 , 628 ' 630 ' 632 ' 634 ' 710 '720, 730, 740, 750, 760, 770, 772, 1010, 1020, 1030, 1040, 1050, 1060, 1070, 1072 '1080: Step HfmBase[n].lstOfs: reference symbol index value HfmBase[n]. lstCode: reference Huffman code Lookup[m']: Huffman code index value

Lookup[m]: reference Huffman code index value

Node : node

Bit : bit

NodeSE : indicator value

Flag : Flag value

Index, Context: field 34

Claims (1)

1343192 J TW5226PA VII. Patent application scope: 1. A decoding method; including: , according to the - bit string - leading bit! The number is obtained from the mth lookup index value of the newly-inquired reference, and the first fixed reference index value is equal to the number of leading bits of the bit string to be decoded; and a reduced Huffman table is obtained according to the first lookup index value The nth reference Huffman code index value, the nth reference Huffman code to index the nth base of the reduced Huffman table and the 丞+戛夫曼 code and an nth ίίΓΓΓ The reference Huffman code 5 丨 value is equal to the bit length of the nth reference Huffman code minus! Obtaining the n+1th most significant bit of the bit string according to the nth reference Huffman code index value; determining the difference between the most important bit of the first state and the nth reference Huffman Whether the value is less than zero; the neck value, the root rib value, and the n base reference index values generate a symbol index value; a symbol is obtained according to the symbol index value. 2. If you apply for a patent range! The decoding method of the item further includes removing, by the bit string, the n+1 most significant bits of the bit string to generate a decoded branch of the first item of the removed difference patent (4), wherein the difference value is equal to The n+1 most significant bits are decremented by the first. Reference Huffman 4. The decoding method described in the patent application section [the above] wherein the 35 1343192 TW5226PA distinct value is equal to the 11th reference Huffman code minus the (10) most significant bit 5: as claimed The method of exhaustion described in item 1 of the scope includes the value. Right, the difference value is less than zero, and the nth reference Huffman code index is adjusted. 6. The decoding method according to item 5 of the patent application scope, in the step of the Huffman code index value, the ηth is decremented Base Huffman code index value. 7. The decoding method according to claim 5, wherein the nth reference Huffman code index value of the nth reference Huffman code index value is integer. $ is incremented by 4 8 · As described in claim 1 of the patent scope, the index value of the basin is equal to the difference value plus the index of the nth reference symbol index / the decoding described in item 1 of the patent scope The method wherein the tiger index value is equal to the nth reference symbol index value minus the difference value. According to two:; Huo's == f decoding method 'the root of the important bit two =; code - the most based on the nth reference Huffman code (four) value to obtain the nth reference Huffman code The length of the bit; and the method for obtaining the _ most significant bit of the bit string according to the length of the bit, as described in the scope of claim 2, wherein the step of taking a symbol is based on The symbol index value is obtained in one of the first gears and in the symbol array. Field 36 TW5226PA ‘ Scope 1 – includes: Optimum Hoffman Table; and = 立 corresponding to the condensed Hoffman table of this new reading. The decoding described in item 12 of the second = range, wherein each step of the bit-count table, is based on a symbol array, an array of array numbers, and the Huffman number reference Huffman code. The reference symbol index value of the Fuman table and the decoding method described in Item 13 of Jian Jianwei, wherein n code values of the nth reference 霍ΙΑ step 'the condensed Huffman table (4)); The length of all the bits in the Mann table is = the decoding method as described in the application _ 13 items. (1) The second step = the simple step = the step of the simplification of the Huffman table. a step in which the value is equal to two m, the symbol represented by the _ datum reference index reference Huffman code is in the symbol array 17. The decoding method as described in claim 12 of the patent application, 1 The step of the newly created lookup table of the Huffman table, the nth reference Huffman code of the nth reduced Huffman table and the reference "Manmarf: the index value generates the newly created lookup table , 37 U4J192 TW5226PA, I8. For the decoding method described in the 12th item of the scope of the application, wherein = construction, corresponding to the step of the newly created lookup table of the simplified Huffman table, the meaning of the streamlined Hough The nth reference Huffman code of the Man table is preceded by (10) the mth reference Huffman code of the newly created lookup table The index value, the nth reference Huffman code index value of the Huffman table. The decoding method according to claim 12, wherein the 6-Hail simplified Huffman table includes a complex number The number of pieces of material is equal to the reference of m. The number of the leading numbers is equal to, the reference Huffman code is corresponding to the index of the Huffman code of the reference, the @ benchmark of the newly created lookup table The Fraunmann code index value is equal to the smallest among the reference Huffman code index values. The decoding method described in claim 12, wherein the ^ hl lite Huffman table includes a plurality of leading numbers equal to The base of the m ho·man code, the (four) guide! The number of Km base Huffman codes respectively correspond to \ plural reference Huffman code index value 'the mth base of the new lookup table' Huffman code The index value is equal to the most = 0 of the reference Huffman code index values 21. The decoding method of claim 12, wherein the mth lookup index value of the newly created lookup table cannot be obtained from the reduced Huffman table. The mth view of the newly created lookup table The index value is equal to the m-1th lookup index value of the new lookup table. ♦ 22. The decoding method of claim 12, wherein the mth lookup (four) value of the newly created lookup table cannot be obtained from the reduced Huffman table by the (7) check. (4) The reference value is equal to the m+1th index value of the newly created lookup table. 38